Shangchen Zhou

The sample in Figure 1 is from the CelebA-HQ, while we train the codebook and CodeFormer only on FFHQ. CodeFormer actually has good generalizability for different datasets, even for real...

> thank you for explanation! One more question, the Fig1(b), what are the six classes of the t-sne clustering results? I guess one of them is the eyes or right...

Hi @FlowDownTheRiver, thanks for your good suggestions! 1. Yes, this project will include an image enhancement model for the background enhancement in future releases, it's actually our main plan. Since...

Hi @FlowDownTheRiver, I updated the face detector, for now our repo supports 4 detection networks, including: - large models: 'YOLOv5l' 'retinaface_resnet50' - small models: 'YOLOv5n' 'retinaface_mobile0.25' We set 'YOLOv5l' by...

@caacoe @FlowDownTheRiver That's awesome to use CodeFormer to enhance AI-created faces. Thanks for your feedback, I could change retina model as the default detection model, and will provide an argument...

@FlowDownTheRiver well noted, I could try to make it in these few days, but I can't make sure since I've been on another project recently.

Hi @FlowDownTheRiver @caacoe, the repo has updated: - Switch 'retinaface_resnet50' as the default face detection model. - Fix the issue of bounding box lines in the face and background blending....

> I'm interested in the new option for BG upscaling via realesrgan, however when I use the flag --bg_upsampler realesrgan, results are identical in comparison without this flag, processing time...

Thanks. We are planning to train a new CodeFormer model with less identity loss, but the downside is that the robustness may be weakened at a cost. You can also...

Hi all @zhangyunming @lazy-nurd, sorry that we cannot make our training code released currently due to some copyright issues. We have to delay our plan to release the training code...